Onix Follows in OMM's Footsteps

In spite of having raised $77.5 million only a few months ago, Onix Microsytems Inc. yesterday scaled back its operations in response to the continuing gloomy outlook for startups making optical switching subsystems.

The company put development of 3D MEMS (micro-electro-mechanical systems) on the backburner and laid off 25 percent of its 150 workers. CEO Steve Panyko says the market for 3D MEMS subsystems has "shifted to the right by one or two years" and, as a result, it didn't make any sense to carry on developing a product of this type at the moment. Instead, Onix will focus its efforts on developing 2D MEMS subsystems.

In doing this, Onix is following in the footsteps of OMM Inc., which shelved its 3D MEMS developments and shed 100 employees a month ago (see OMM-inous News).

In both cases, the fundamental problem is that it's very tough to know when the market for their optical switching modules will revive, partly because the startups sell to system vendors -- not to the service providers that generate demand for their products. This applies to 2D as well as 3D MEMS subsystems, although startups are loath to acknowledge this.

OMM is further ahead than Onix in developing 2D MEMS modules. It's shipping them, while Onix's modules are still under development. All the same, Onix figures it's got some advantages over OMM on two counts -- technology and manufacturing strategy. "It's not always the first to market that ends up being one of the winners in the long term," notes Meng-Hsuing Kiang, Onix's founder and director of business development.

Onix uses electomagnetic force to turn its arrays of tiny tilting mirrors, while OMM uses much weaker electrostatic forces. Electromagnetic systems have much lower power requirements, according to Kiang. Conrad Burke, OMM's senior VP of marketing and business development counters that OMM chose electrostatic technology "because it promised to make a more reliable and robust product."

Kiang also makes a big thing out of Onix's strategy of outsourcing the manufacturing of its MEMS fabric to Analog Devices Inc. (NYSE: ADI). It's using a similar strategy for packaging its modules, outsourcing the task to a heavyweight company in that field. OMM's Burke calls this is recipe for problems: "The technology isn't outsourceable," he says, contending that having manufacturing expertise in house is essential.

— Peter Heywood, Founding Editor, Light Reading
HarveyMudd 12/4/2012 | 7:52:38 PM
re: Onix Follows in OMM's Footsteps ONIX does not seem to have any future. Its management team is very week. The VC fuding the company did not know it either what they were getting into, but it was someone else's money. VCs do not care if they are funding with money from other people.
daylight 12/4/2012 | 7:52:31 PM
re: Onix Follows in OMM's Footsteps That is a very broad and derogatory statement to make. Do you have evidence to back it up? Or are you just flaming Onix for no reason?
iamnoone 12/4/2012 | 7:52:26 PM
re: Onix Follows in OMM's Footsteps I think Onix does have some good technology compared to OMM, using single crystal silicon instead of polysilicon, and using torsional bars instead of joints that are prone to friction. I don't know the reliability of these torsional bars that need to twist very large angles (90 degrees?), though. Beware of reliability studies that show many millions of cycles -- MEMS devices survive constant cycling much better than switch-and-hold. The need for large angle rotation is probably why the article emphasizes the use of stronger electromagnetic forces rather than electrostatic forces (which probably cannot work in Onix's design).

iamnoone 12/4/2012 | 7:52:25 PM
re: Onix Follows in OMM's Footsteps Maybe I should have clarified my point about polysilicon: it's not the reliability of polysilicon per se that makes single crystal silicon a better choice -- it's the ability to make it flat to less than a quarter-wavelength over mirror diameters of 1mm or more. There's been a lot of work on making flat polysilicon, but single crystal silicon comes out flat and with little surface roughness without any additional work. Hence the possibility of higher yield. The ability to deposit metal and still keep it flat after post processing and environmental testing is another issue.

Once again, the comparison between airbag sensors and MEMS mirrors is misleading because the range of motion in sensors is orders of magnitude smaller (sub-micron vs tens of microns). The strains that are seen are much bigger in moving mirrors. Trillions of cycles on Analog Devices MEMS sensors are certainly true, but even millions of cycles with a 1-hour hold with 10 to 90 degree rotation has never been tested, although I'm not saying it cannot be achieved. N.B. there are only about 9000 hrs in a year. Cyclic fatigue is quite different from static fatigue.
MEMS guy 12/4/2012 | 7:52:25 PM
re: Onix Follows in OMM's Footsteps Your Point:
"but even millions of cycles with a 1-hour hold with 10 to 90 degree rotation has never been tested, although I'm not saying it cannot be achieved."

Just because it has not been published, does not mean it has not been done:-)

Strain in moving mirrors is a function of the design . . . straightforward designs can be made that keep you orders of magnitude away from trouble spots, even for large tilt angles.

While bulk mirrors are easier to make flat, they generate their own set of challenges. Single crystal is not necessarily more intrinsically flat than what is achievable in polycrystalline silicon. The major issue is that people make very thick bulk mirrors that are very stiff. This is the more dominant effect than poly vs. single crystal. Heavy stiff mirrors with flimsy suspensions lead to problems with shock, and heavy mirrors on beefy springs lead to challenges in getting reasonable tilt angles. Going to an exotic actuation scheme to overcome this can then lead to manufacturing problems.

In my previous post I acknowledged that accelerometers are much simpler designs than optical switches, but there is still relevant things that can be learned . . . It shows you have a technology that at its core is manufacturable, affordable, and some of the qualification issues are similar (stiction, for example, can be effectively dealt with)

My point is that there are many tradeoffs between performance, manufacturability, and reliability.

Whether Bulk or Surface wins out in the end will be a function of how well any given group executes within their selected technology. Simply saying bulk is better because it is easier to get flatness overlooks a world of issues. What is required is a strong team, good management, good financial backing, and solid manufacturing experience/infrastructure.
MEMS guy 12/4/2012 | 7:52:25 PM
re: Onix Follows in OMM's Footsteps I am not associated with OMM. I disagree with your reliability analysis. You imply that there is some intrinsic problem with polycrystalline silicon. Despite all the bru ha ha over the issue, the fact remains that OMM is further along than most with telcordia qualification of MEMS systems. In addition, virtually 100% of cars being built today use surface mems for air bag deployment. While the optical switch designs are much more complicated, the core technology and material set is very similar.

There is no data to suggest that polysilicon has a problem with switch and hold. There has been experiments amasing TRILLIONS of cycles on polysilicon devices with no detectable degradation, as well as extensive long term tests for sample and hold. In fact, the airbag applicaiton operates the polysilicon devices in a "hold" architecture. When designed properly, surface MEMS can have virtually infinite lifetime.

I do agree with your point on friction . . . if not designed properly, friction, stiction and wear can be a formidable challenge in any MEMS technology.

The key is proper design, fabrication, and packaging. If reliability is taken into account up-front in these three areas, polycrystalline MEMS offer a mature, reliable solution.

Will everyone be successful . . . no. Are the issues managable if you know what you are doing . . . yes.

Sign In